A combination of Model Output Statistics (MOS) forecasts from the ETA and AVN forecast models and raw model output from the Short Range Ensemble Forecast (SREF) was used to produce 18-42 hour forecasts of precipitation amount, maximum temperature, and minimum temperature. This combination product, dubbed ?BJEGUI?, participated in the National Collegiate Weather Forecasting Contest during the 2002-2003 academic year. BJEGUI had smaller forecast errors than 94% of all human forecasters that qualified, as well as smaller forecast errors than other routinely available guidance products. As a stand alone forecast, a bias corrected SREF forecast had comparable skill to MOS products in predicting maximum and minimum temperature. SREF data was also a beneficial component to BJEGUI, resulting in better forecasts than a combination of ETA MOS and AVN MOS only. An optimal blending of the three components, ETA MOS, AVN MOS, and SREF, per forecast parameter per forecast site was applied. This blending resulted in better forecasts than a straight average of the three components.

A first attempt at a new post-processing scheme, weighting the SREF ensemble forecasts using the Ensemble Transform Kalman Filter (ET KF), is presented. The goal of the ET KF weighting is to replace a pure mean of the ensemble members with a forecast which more heavily weights the ensemble members that best match observations taken at some set time between ensemble generation and forecast verification time. Results from this ET KF weighting were inconclusive, highlighting the need for more research on this method.